A swarm appeared again a couple of days later, he said, and by then it was much better organized. "It displayed distinctive swarming behavior, that sort of swirling in the cloud that you've seen. So it was clear that it was our stuff."
"And what happened then?"
"The swarm swirled around the desert near the installation, like before. It came and went. For the next few days, we tried to gain control of it by radio, but we never could. And eventually-about a week after that-we found that none of the cars would start." He paused. "I went out there to have a look, and I found that all the onboard computers were dead. These days all automobiles have microprocessors built into them. They control everything from fuel injection to radios and door locks."
"But now the computers were not functioning?"
"Yeah. Actually, the processor chips themselves were fine. But the memory chips had eroded. They'd literally turned to dust."
I thought, Oh shit. I said, "Could you figure out why?"
"Sure. It wasn't any big mystery, Jack. The erosion had the characteristic signature of gamma assemblers. You know about that? No? Well, we have nine different assemblers involved in manufacturing. Each assembler has a different function. The gamma assemblers break down carbon material in silicate layers. They actually cut at the nano level-slicing out chunks of carbon substrate."
"So these assemblers cut the memory chips in the cars."
"Right, right, but ..." David hesitated. He was acting as if I were missing the point. He tugged at his cuffs, fingered his collar. "The thing you have to keep in mind, Jack, is that these assemblers can work at room temperature. If anything, the desert heat's even better for them. Hotter is more efficient."
For a moment I didn't understand what he was talking about. What difference did it make about room temperature or desert heat? What did that have to do with memory chips in cars? And then suddenly, finally, the penny dropped.
"Holy shit," I said.
He nodded. "Yeah."
David was saying that a mixture of components had been vented into the desert, and that these components-which were designed to self-assemble in the fabrication structure-would also self-assemble in the outside world. Assembly could be carried out autonomously in the desert. And obviously, that's exactly what was happening.
I ticked the points off to make sure I had it right. "Basic assembly begins with the bacteria. They've been engineered to eat anything, even garbage, so they can find something in the desert to live off of."
"Which means the bacteria multiply, and begin churning out molecules that self-combine, forming larger molecules. Pretty soon you have assemblers, and the assemblers begin to do the final work and turn out new microagents."
"Which means that the swarms are reproducing."
"Yes. They are."
"And the individual agents have memory."
"Yes. A small amount."
"And they don't need much, that's the whole point of distributed intelligence. It's collective. So they have intelligence, and since they have memory, they can learn from experience."
"And the PREDPREY program means they can solve problems. And the program generates enough random elements to let them innovate."
My head throbbed. I was seeing all the implications, now, and they weren't good. "So," I said, "what you're telling me is this swarm reproduces, is self-sustaining, learns from experience, has collective intelligence, and can innovate to solve problems."
"Which means for all practical purposes, it's alive."
"Yes." David nodded. "At least, it behaves as if it is alive. Functionally it's alive, Jack."
I said, "This is very fucking bad news."
Brooks said, "Tell me."
"I'd like to know," I said, "why this thing wasn't destroyed a long time ago."
David said nothing. He just smoothed his tie, and looked uncomfortable. "Because you realize," I said, "that you're talking about a mechanical plague. That's what you've got here. It's just like a bacterial plague, or a viral plague. Except it's mechanical organisms. You've got a fucking man-made plague."
He nodded. "Yes."
"And it's not limited by biological rates of evolution. It's probably evolving much faster."
He nodded. "It is evolving faster."
"How much faster, David?"
Brooks sighed. "Pretty damn fast. It'll be different this afternoon, when it comes back."
"Will it come back?"
"It always does."
"And why does it come back?" I said.
"It's trying to get inside."
"And why is that?"
David shifted uncomfortably. "We have only theories, Jack."
"One possibility is that it's a territorial thing. As you know, the original PREDPREY code includes a concept of a range, of a territory in which the predators will roam. And within that core range, it defines a sort of home base, which the swarm may consider to be the inside of this facility."
I said, "You believe that?"
"Not really, no." He hesitated. "Actually," he said, "most of us think that it comes back looking for your wife, Jack. It's looking for Julia."
That was how, with a splitting headache, I found myself on the phone to the hospital in San Jose. "Julia Forman, please." I spelled the name for the operator. "She's in the ICU," the operator said.
"Yes, that's right."
"I'm sorry but direct calls are not allowed."
"Then the nursing station."
"Thank you, please hold."
I waited. No one was answering the phone. I called back, went through the operator again, and finally got through to the ICU nursing station. The nurse told me Julia was in X-ray and didn't know when she would be back. I said Julia was supposed to be back by now. The nurse said rather testily that she was looking at Julia's bed right now, and she could assure me Julia wasn't in it.
I said I'd call back.
I shut the phone and turned to David. "What was Julia doing in all this?"
"Helping us, Jack."
"I'm sure. But how, exactly?"
"In the beginning, she was trying to coax it back," he said. "We needed the swarm close to the building to take control again by radio. So Julia helped us keep it close."
"Well, she entertained it."
"I guess you'd call it that. It was very quickly obvious that the swarm had rudimentary intelligence. It was Julia's idea to treat it like a child. She went outside with bright blocks, toys. Things a kid would like. And the swarm seemed to be responding to her. She was very excited about it."
"The swarm was safe to be around at that time?"
"Yes, completely safe. It was just a particle cloud." David shrugged. "Anyway, after the first day or so, she decided to go a step further and formally test it. You know, test it like a child psychologist."
"You mean, teach it," I said.
"No. Her idea was to test it."
"David," I said. "That swarm's a distributed intelligence. It's a goddamn net. It'll learn from whatever you do. Testing is teaching. What exactly was she doing with it?"
"Just, you know, sort of games. She'd lay out three colored blocks on the ground, two blue and one yellow, see if it would choose the yellow. Then with squares and triangles. Stuff like that."
"But David," I said. "You all knew this was a runaway, evolving outside the laboratory. Didn't anybody think to just go out and destroy it?"
"Sure. We all wanted to. Julia wouldn't allow it."
"She wanted it kept alive."
"And nobody argued with her?"
"She's a vice president of the company, Jack. She kept saying the swarm was a lucky accident, that we had stumbled onto something really big, that it could eventually save the company and we mustn't destroy it. She was, I don't know, she was really taken with it. I mean, she was proud of it. Like it was her invention. All she wanted to do was 'rein it in.' Her words."
"Yeah. Well. How long ago did she say that?"
"Yesterday, Jack." David shrugged. "You know, she only left here yesterday afternoon." It took me a moment to realize that he was right. Just a single day had passed since Julia had been here, and then had had her accident. And in that time, the swarms had already advanced enormously.
"How many swarms were there yesterday?"
"Three. But we only saw two. I guess one was hiding." He shook his head. "You know, one of the swarms had become like a pet to her. It was smaller than the others. It'd wait for her to come outside, and it always stuck close to her. Sometimes when she came out it swirled around her, like it was excited to see her. She'd talk to it, too, like it was a dog or something." I pressed my throbbing temples. "She talked to it," I repeated. Jesus Christ. "Don't tell me the swarms have auditory sensors, too."
"No. They don't."
"So talking was a waste of time."
"Uh, well ... we think the cloud was close enough that her breath deflected some of the particles. In a rhythmic pattern."
"So the whole cloud was one giant eardrum?"
"In a way, yeah."
"And it's a net, so it learned ..."
I sighed. "Are you going to tell me it talked back?"
"No, but it started making weird sounds."
I nodded. I'd heard those weird sounds. "How does it do that?"
"We're not sure. Bobby thinks it's the reverse of the auditory deflection that allows it to hear. The particles pulse in a coordinated front, and generate a sound wave. Sort of like an audio speaker."
It would have to be something like that, I thought. Even though it seemed unlikely that it could do it. The swarm was basically a dust cloud of miniature particles. The particles didn't have either the mass or the energy to generate a sound wave.
A thought occurred to me. "David," I said, "was Julia out there yesterday, with the swarms?"
"Yes, in the morning. No problem. It was a few hours later, after she left, that they killed the snake."
"And was anything killed before that?"
"Uh ... possibly a coyote a few days ago, I'm not sure."
"So maybe the snake wasn't the first?"
"And today they killed a rabbit."
"Yeah. So it's progressing fast, now."
"Thank you, Julia," I said.
I was pretty sure the accelerated behavior of the swarms that we were seeing was a function of past learning. This was a characteristic of distributed systems-and for that matter a characteristic of evolution, which could be considered a kind of learning, if you wanted to think of it in those terms. In either case, it meant that systems experienced a long, slow starting period, followed by ever-increasing speed.
You could see that exact speedup in the evolution of life on earth. The first life shows up four billion years ago as single-cell creatures. Nothing changes for the next two billion years. Then nuclei appear in the cells. Things start to pick up. Only a few hundred million years later, multicellular organisms. A few hundred million years after that, explosive diversity of life. And more diversity. By a couple of hundred million years ago there are large plants and animals, complex creatures, dinosaurs. In all this, man's a latecomer: four million years ago, upright apes. Two million years ago, early human ancestors. Thirty-five thousand years ago, cave paintings. The acceleration was dramatic. If you compressed the history of life on earth into twenty-four hours, then multicellular organisms appeared in the last twelve hours, dinosaurs in the last hour, the earliest men in the last forty seconds, and modern men less than one second ago. It had taken two billion years for primitive cells to incorporate a nucleus, the first step toward complexity. But it had taken only 200 million years-one-tenth of the time-to evolve multicellular animals. And it took only four million years to go from small-brained apes with crude bone tools to modern man and genetic engineering. That was how fast the pace had increased.
This same pattern showed up in the behavior of agent-based systems. It took a long time for agents to "lay the groundwork" and to accomplish the early stuff, but once that was completed, subsequent progress could be swift. There was no way to skip the groundwork, just as there was no way for a human being to skip childhood. You had to do the preliminary work. But at the same time, there was no way to avoid the subsequent acceleration. It was, so to speak, built into the system.
Teaching made the progression more efficient, and I was sure Julia's teaching had been an important factor in the behavior of the swarm now. Simply by interacting with it, she had introduced a selection pressure in an organism with emergent behavior that couldn't be predicted. It was a very foolish thing to do.
So the swarm-already developing rapidly-would develop even more rapidly in the future. And since it was a man-made organism, evolution was not taking place on a biological timescale. Instead, it was happening in a matter of hours.
Destroying the swarms would be more difficult with each passing hour. "Okay," I said to David. "If the swarms are coming back, then we better get ready for them." I got to my feet, wincing at the headache, and headed for the door. "What do you have in mind?" David said.
"What do you think I have in mind?" I said. "We've got to kill these things cold stone dead. We have to wipe them off the face of the planet. And we have to do it right now." David shifted in his chair. "Fine with me," he said. "But I don't think Ricky's going to like it."
David shrugged. "He's just not."
I waited, and said nothing.
David fidgeted in his chair, more and more uncomfortable. "The thing is, he and Julia are, uh, in agreement on this."
"They're in agreement."
"Yes. They see eye to eye. I mean, on this."
I said, "What are you trying to say to me, David?"
"Nothing. Just what I said. They agree the swarms should be kept alive. I think Ricky's going to oppose you, that's all."
I needed to talk to Mae again. I found her in the biology lab, hunched over a computer monitor, looking at images of white bacterial growth on dark red media. I said, "Mae, listen, I've talked to David and I need to-uh, Mae? Have you got a problem?" She was looking fixedly at the screen.
"I think I do," she said. "A problem with the feedstock."
"What kind of problem?"
"The latest Theta-d stocks aren't growing properly." She pointed to an image in the upper corner of the monitor, which showed bacteria growing in smooth white circles. "That's normal coliform growth," she said. "That's how it's supposed to look. But here ..." She brought up another image in the center of the screen. The round forms appeared moth-eaten, ragged and misshapen. "That's not normal growth," she said, shaking her head. "I'm afraid it's phage contamination."
"You mean a virus?" I said. A phage was a virus that attacked bacteria. "Yes," she said. "Coli are susceptible to a very large number of phages. T4 phage is of course the most common, but Theta-d was engineered to be T4-resistant. So I suspect it's a new phage that's doing this."
"A new phage? You mean it's newly evolved?"
"Yes. Probably a mutant of an existing strain, that somehow gets around the engineered resistance. But it's bad news for manufacturing. If we have infected bacterial stocks, we'll have to shut down production. Otherwise we'll just be spewing viruses out."
"Frankly," I said, "shutting down production might be a good idea."
"I'll probably have to. I'll try to isolate it, but it looks aggressive. I may not be able to get rid of it without scrubbing the kettle. Starting over with fresh stock. Ricky's not going to like it."
"Have you told him about this?"
"Not yet." She shook her head. "I don't think he needs more bad news right now. And besides ..." She stopped, as if she had thought better of what she was going to say. "Besides what?"
"Ricky has a huge stake in the success of this company." She turned to face me. "Bobby heard him on the phone the other day, talking about his stock options. And sounding worried. I think Ricky sees Xymos as his last big chance to score. He's been here five years. If this doesn't work out, he'll be too senior to start over at a new company. He's got a wife and baby; he can't gamble another five years, waiting to see if the next company clicks. So he's really trying to make this happen, really driving himself. He's up all night, working, figuring. He isn't sleeping more than three or four hours. Frankly, I worry it's affecting his judgment."
"I can imagine," I said. "The pressure must be terrible."
"He's so sleep-deprived it makes him erratic," Mae said. "I'm never sure what he'll do, or how he'll respond. Sometimes I get the feeling he doesn't want to get rid of the swarms at all. Or maybe he's scared."
"Maybe," I said.
"Anyway, he's erratic. So if I were you I'd be careful," she said, "when you go after the swarms. Because that's what you're going to do, isn't it? Go after them?"
"Yes," I said. "That's what I'm going to do."
They had all gathered in the lounge, with the video games and pinball machines. Nobody was playing them now. They were watching me with anxious eyes as I explained what we had to do. The plan was simple enough-the swarm itself was dictating what we had to do, although I was skipping that uncomfortable truth.
Basically, I told them we had a runaway swarm we couldn't control. And the swarm exhibited self-organizing behavior. "Whenever you have a high SO component, it means the swarm can reassemble itself after an injury or disruption. Just as it did with me. So this swarm has to be totally, physically destroyed. That means subjecting the particles to heat, cold, acid, or high magnetic fields. And from what I've seen of its behavior, I'd say our best chance to destroy it is at night when the swarm loses energy and sinks to the ground."
Ricky whined, "But we already told you, Jack, we can't find it at night-"
"That's right, you can't," I said. "Because you didn't tag it. Look, it's a big desert out there. If you want to trace it back to its hiding place, you've got to tag it with something so strong you can follow its trail wherever it goes."
"Tag it with what?"
"That's my next question," I said. "What kind of tagging agents have we got around here?" I was greeted with blank looks. "Come on, guys. This is an industrial facility. You must have something that will coat the particles and leave a trail we can follow. I'm talking about a substance that fluoresces intensely, or a pheromone with a characteristic chemical signature, or something radioactive ... No?"
More blank looks. Shaking their heads.
"Well," Mae said, "of course, we have radioisotopes."
"All right, fine." Now we were getting somewhere.
"We use them to check for leaks in the system. The helicopter brings them out once a week."
"What isotopes do you have?"
"Selenium-72 and Rhenium-186. Sometimes Xenon-133 as well. I'm not sure what we've got on hand right now."
"What kind of half-lives are we talking about?" Certain isotopes lost radioactivity very rapidly, in a matter of hours or minutes. If so, they wouldn't be useful to me. "Half-life averages about a week," Mae said. "Selenium's eight days. Rhenium's four days. Xenon-133 is five days. Five and a quarter."
"Okay. Any of them should do fine for our purposes," I said. "We only need the radioactivity to last for one night, after we tag the swarm."
Mae said, "We usually put the isotopes in FDG. It's a liquid glucose base. You could spray it."
"That should be fine," I said. "Where are the isotopes now?"
Mae smiled bleakly. "In the storage unit," she said.
"Where is that?"
"Outside. Next to the parked cars."
"Okay," I said. "Then let's go out and get them."
"Oh, for Christ's sake," Ricky said, throwing up his hands. "Are you out of your mind? You nearly died out there this morning, Jack. You can't go back out."
"There isn't any choice," I said.
"Sure there is. Wait until nightfall."
"No," I said. "Because that means we can't spray them until tomorrow. And we can't trace and destroy them until tomorrow night. That means we wait thirty-six hours with an organism that is evolving fast. We can't risk it."
"Risk it? Jack, if you go out now, you'll never survive. You're fucking crazy even to consider it."
Charley Davenport had been staring at the monitor. Now he turned to the group. "No, Jack's not crazy." He grinned at me. "And I'm going with him." Charley began to hum: "Born to Be Wild."
"I'm going, too," Mae said. "I know where the isotopes are stored."
I said, "It's not really necessary, Mae, you can tell me-"
"No. I'm coming."
"We'll need to improvise a spray apparatus of some kind." David Brooks was rolling up his sleeves carefully. "Presumably, remotely controlled. That's Rosie's specialty."
"Okay, I'll come, too," Rosie Castro said, looking at David. "You're all going?" Ricky stared from one to another of us, shaking his head. "This is extremely risky," he said. "Extremely risky."
Nobody said anything. We all just stared at him.
Then Ricky said, "Charley, will you shut the fuck up?" He turned to me. "I don't think I can allow this, Jack ..."
"I don't think you have a choice," I said.
"I'm in charge here."
"Not now," I said. I felt a burst of annoyance. I felt like telling him he'd screwed the pooch by allowing a swarm to evolve in the environment. But I didn't know how many critical decisions Julia had made. In the end, Ricky was obsequious to management, trying to please them like a child pleasing a parent. He did it charmingly; that was how he had moved ahead in life. That was also his greatest weakness.
But now Ricky stuck out his chin stubbornly. "You just can't do it, Jack," he said. "You guys can't go out there and survive."
"Sure we can, Ricky," Charley Davenport said. He pointed to the monitor. "Look for yourself."
The monitor showed the desert outside. The early afternoon sun was shining on scrubby cactus. One stunted juniper in the distance, dark against the sun. For a moment I didn't understand what Charley was talking about. Then I saw the sand blowing low on the ground. And I noticed the juniper was bent to one side.
"That's right, folks," Charley Davenport said. "We got a high wind out there. High wind, no swarms-remember? They have to hug the ground." He headed toward the passageway leading to the power station. "Time's a-wasting. Let's do it, guys." Everybody filed out. I was the last to leave. To my astonishment, Ricky pulled me aside, blocked the door with his body. "I'm sorry, Jack, I didn't want to embarrass you in front of the others. But I just can't let you do this."
"Would you rather have somebody else do it?" I said.
He frowned. "What do you mean?"
"You better face facts, Ricky. This is already a disaster. And if we can't get it under control right away, then we have to call for help."
"Help? What do you mean?"
"I mean, call the Pentagon. Call the Army. We have to call somebody to get these swarms under control."
"Jesus, Jack. We can't do that."
"We have no choice."
"But it would destroy the company. We'd never get funding again."
"That wouldn't bother me one bit," I said. I was feeling angry about what had happened in the desert. A chain of bad decisions, errors and fuckups extending over weeks and months. It seemed as if everyone at Xymos was doing short-term solutions, patch-and-fix, quick and dirty. No one was paying attention to the long-term consequences.
"Look," I said, "you've got a runaway swarm that's apparently lethal. You can't screw around with this anymore."
"Julia isn't here."
"But she said-"
"I don't care what she said, Ricky."
"But the company-"
"Fuck the company. Ricky." I grabbed him by the shoulders, shook him once hard. "Don't you get it? You won't go outside. You're afraid of this thing, Ricky. We have to kill it. And if we can't kill it soon, we have to call for help."
"We'll see about that," he snarled. His body tensed, his eyes flared. He grabbed my shirt collar. I just stood there, staring at him. I didn't move. Ricky glared at me for a moment, and then released his grip. He patted me on the shoulder and smoothed out my collar. "Ah hell, Jack," he said. "What am I doing?" And he gave me his self-deprecating surfer grin. "I'm sorry. I think the pressure must be getting to me. You're right. You're absolutely right. Fuck the company. We have to do this. We have to destroy those things right away."
"Yes," I said, still staring at him. "We do."
He paused. He took his hand away from my collar. "You think I'm acting weird, don't you? Mary thinks I'm acting weird, too. She said so, the other day. Am I acting weird?"